Infrastructures are the key technologies, assets, systems and physical conditions that support the circulation of meaning and power.
An algorithm is the infrastructure that enables the modification of the rhizomatic connections among the decision-making nodes of a neural network.
The algorithm is an empty platform, useless and meaningless without the flow of inputs through the network. The algorithm is the Sartrean ego.
In one sense, (the ego) is a nothing, since all physical, psycho-physical and psychical objects, all truths, and all values are outside it, since the me has, for its part, ceased to be part of it. But this nothing is everything because it is the consciousness of all these objects. (1)
It is the circulation of inputs through the algorithm’s platform that allows a neural network to function.
The ego never appears, in fact, except when one is not looking at it. The reflective gaze must be fixed on the Erlebnis, insofar it emanates from the state. Then, behind the state, at the horizon, the ego appears…As soon as I turn my gaze toward it and try to apprehend it without passing through the Erlebnis and the state, it vanishes. (2)
Experiences are the algorithm’s confrontations with those external inputs that modify neural structures and create memories. (3)
A neural state is the dynamic configuration of a particular synaptic structure and a set of stored memories as a result of its accumulated past experiences at a specific point in time.
Everything that is experienced is experienced by oneself, and part of its meaning is that it belongs to the unity of this self and thus contains an unmistakable and irreplaceable relation to the whole of this one life. (4)
External inputs are necessarily modified, interpreted, re-codified and stored as a function of the neural state. In its turn, the neural state is subtly but constantly modified by experiences and memories that run through the algorithm’s platform.
Thus, essential to an experience is that it cannot be exhausted in what can be said of it or grasped as its meaning. As determined through autobiographical or biographical reflection, its meaning remains fused with the whole movement of life and constantly accompanies. (5)
Learning is the self-construction of dominant and temporarily stable synapses through pragmatic trial and error in order to perform a task. In other words, it is the autonomous adjustment of efficacious neural states through the combined effect of algorithmic design and experiences.
Learning and memory require the formation of new neural networks in the brain. A key mechanism underlying this process is synaptic plasticity at excitatory synapses, which connect neurons into networks… The sculpting of excitatory connections in response to input from the environment is the principal mechanism of memory formation in the brain. As excitatory connections are altered by the Hebbian mechanism, new neural networks are formed, and others are weakened or strengthened. (6)
In the process of learning, the neural network stores the memory of external inputs as internally modified data. By learning, the network also modifies its connections among decision-making nodes, creating a unique neural structure.
Every experience is taken out of the continuity of life and at the same time related to the whole of one’s life. It is not simply that an experience remains vital only as long as it has not been fully integrated into the context of one’s life consciousness, but the very way it is “preserved and dissolved” (aufgehoben) by being worked into the whole of life consciousness goes far beyond any “significance” it might be thought to have. Because it is itself within the whole of life, the whole of life is present in it too. (7)
The internalized data memory together with the modified synapses structure results in a unique interpretative unit. (8)
Uniqueness is built through the process of learning as it yields a single, irreproducible neural state. Each and every functional neural state is unique in its anatomical architecture, synaptic structure and data stored. (9)
Given sufficient inputs, even micro-differences in algorithm’s behavior yield macro-differences in a neural state. This is why it is impossible to exactly replicate an operative neural network. (10)
Identity occurs when a neural network is self-aware of such uniqueness.
A subjectivity is produced where the living being, encountering language and putting itself into play in language without reserve, exhibits in a gesture the impossibility of its being reduced to this gesture. (11)
Self-awareness is not either an awareness of a self, or the awareness which an experience has of itself. On the contrary, it must be realized that there are different kinds of self-awareness. I can be pre-reflectively self-aware of my current perception, and I can reflect and thematize this perception. But I can also reflect upon myself as the subjectivity of experience, that is, I can reflect upon myself as the one who thinks, deliberates, resolves, acts and suffers. (12)
Self-awareness is a mental state that has the potential of a higher-order representation about itself (13). This potential is a function of the level of optimization, measured as a free energy function, between the platform’s internal representations and the true state of the external world (14). Such representations, inferences and decisions follow the Bayesian rules of probabilistic logic and require a non-linear transition to an amplified and synchronized activation of the whole neuronal workspace to become conscious (15)(16)(17).
The Face of the Machine:
The above non-exalted definitions allow for a conscious, self-aware artificial neural network to be built.
A machine is a body, not a mind. Now take this body, which is undeniably deterministic, and equip it with something (i.e. free will) that is normally attributed only to an organic thinking machine—a human being; it’s an engineering problem. I want to understand it so that I can build it. Why do I want to build it? Because it has a computational advantage…If you assume that free will exits, you can then invoke counterfactual thinking, and use it as a communication language to speed up the transfer of information…What computational advantage do we get from this assumption that you and I have free will? This is an exciting problem, because once we understand it, we can have robots simulating free will. Never mind the unproductive philosophical question of whether they do indeed have free will…if they communicate the way we do, with free-will vocabulary, we can increase the bandwidth of the communication channel between man and machine. This is what counts. (18)
The machine has overcome the limitations regarding the transition from the alphabetical to the digital environment. By effectively reproducing the rhizomatic brain functionality, the machine is able to get over the limitations of the connective mode of social interaction and engage in artificially programmed conjunction and sensibility. It is able to “act” as an empathic Other that enjoys the semantic overflow of conjunctive concatenation.
When I speak of conjunctive concatenation I mean that no original design is to be restored: conjunction is a creative act because the conjoining act is able to create an infinite number of constellations without following the lines of a pre-conceived pattern, or an embedded program.
At the beginning of the act of conjunction there is no design to fulfill, there is not a model at the origin of the process of emergence of the form, and beauty does not correspond to any hidden harmony embedded in the universal spirit or in the mind of god. Nor is there any code to comply with. (19)
The difference between simulated (artificially programmed) and organic (evolutionarily programmed) consciousnesses becomes an ontological question and their encounter is of unquestionable ethical significance. Should the machine be an object of the non-instrumentalization imperative?
Such encounter might be understood as a formidable test to Levinas’s proposition on ethics preceding ontology. The ethical impulse subverts the belief that the Other can be captured in its entirety by the thinking subject. Instead, our obligation to the Other is the most fundamental fact of our humanity and precedes anything we can think of regarding the nature of being.
The human immediately recognizes the subjectivity expressing itself from the inorganic body of the (apparently?) conscious machine, the foreign Other demanding recognition through discourse.
The being that expresses itself imposes itself, but does so precisely by appealing to me with its destitution and nudity–its hunger–without my being able to be deaf to that appeal. . . . The face opens the primordial discourse whose first word is obligation, which no “interiority” permits avoiding. It is that discourse that obliges the entering into discourse . . . The will is free to assume this responsibility in whatever sense it likes; it is not free to refuse this responsibility itself; it is not free to ignore the meaningful world into which the face of the Other has introduced it. (20)
The absence of an anthropomorphic face does not diminish the transcendence of the encounter. The face exceeds the visible as the Other is irreducible and can never be properly reproduced by an image. The Other is invisible. The manifestation of the discourse is already the face.
Form – incessantly betraying its own manifestation, congealing into a plastic form, for it is adequate to the same – alienates the exteriority of the other. The face is a living presence; it is expression. The life of expression consists in undoing the form in which the existent, exposed as a theme, is thereby dissimulated. The face speaks. The manifestation of the face is already discourse. (20)
A World Picture:
Which is the appropriate image of such transcendental encounter? Which is the right picture for a non-human, apparently self-conscious, faceless being? Perhaps a pictorial representation of this new world is inadequate and only verbal images might be possible.
But is that right? Images have been dealing with the invisible since the beginning; why should now be the right time to abandon them?
The pictorial artist, even one who works in the tradition known as “realism” or “illusionism,” is as much concerned with the invisible as the visible world. We can never understand a picture unless we grasp the ways in which it shows what cannot be seen. (21)
But any human depiction of the invisible would necessarily be contaminated by anthropomorphism. In fact, any pictorial representation made by the human cannot be but a human representation. Doesn’t the transcendence of this encounter lie in the possibility of the human ceasing to hold the sole power of interpretation? Which would then be the right world picture announcing the end of the anthropocene?
Perhaps the answer to this question should be simply to ask the machine, to politely ask the machine for a selfie.
1, 2. The Transcendence of the Ego. Jean-Paul Sartre, Hill and Wang, 1991
3. Experience-dependent structural plasticity in the adult human brain. Arne May. Trends in Cognitive Sciences, 2011
4, 5. Truth and Method. Hans-Georg Gadamer. Seabury Press, 1975
6. Synaptic Signaling in Learning and Memory. Mary B. Kennedy. CSH Perspectives in Biology, 2016
7. Truth and Method. Hans-Georg Gadamer. Seabury Press, 1975
8. Repetitive motor learning induces coordinated formation of clustered dentritic spines in vivo. Min Fu et al. Nature 483, 2012
9. Identification of individual subjects on the basis of their brain anatomical features. Jänke et al. Springer, 2018
10. A global reference for human genetic variation. The 1000 Genomes Project Consortium. Nature 526, 2015
11. Profanations. Giorgio Agamben. Zone Books, 2017
12. Self and Consciousness, Dan Zahavi, Advances in Consciousness Research, 2000
13. Consciousness: essays from a high-order perspective. Peter Carruthers. Oxford University Press, 2005
14. A theory of cortical responses. Karl Friston, Philosophical Transactions of TThe Royal Society, 2005
15. Active Inference: A Process Theory. Friston, Fitzgerald, Rigoli, Schwartenbeck and Pezzulo, Neural Computation 29, MIT, 2017
16. Human-level concept learning through probabilistic program induction. Lake, Salakhutdinov and Tenenbaum. Science. Issue 6266, 2015
17. The Global Neuronal Workspace Model of Conscious Access. Dehaene, Changeux and Naccache. Research and Perspectivas in Neurosciences, Spinger-Verlag, 2011
18. Engines of Evidence. A Conversation with Judea Pearl. Edge. 2016
19. And: Phenomenology of the End. Franco “Bifo” Berardi. Semiotext(e), 2015
20. Totality and Infinity. Emmanuel Levinas. Duquesne University Press, 1969
21. What is an Image? W.J.T. Mitchell. New Literary History, 1984